This fixes a crash which occurs when the terminal is reset while
searching, due to the vi mode cursor being outside of the visible area.
This also fixes an issue where the search state reset would incorrectly
clamp the vi mode cursor to the grid, rather than the absolute viewport
position.
While this fix does resolve all crashes when searching while running
`cat /dev/urandom`, it does raise the question if manually clamping the
vi mode cursor in every location where it is modified is the right
choice.
A follow-up to provide a safer API which guarantees correct modification
of the vi mode cursor location is probably a good idea.
Fixes#5942.
Currently Alacritty only works on hardware which supports OpenGL 3.3 or
more, which can become problematic with older devices. This patch adds a
new GLES2 renderer, since it is much more widely supported, especially
on weaker hardware like phones or a Raspberry Pi.
While the GLES2 renderer is slower than the OpenGL 3.3+ version, it is
still significantly faster than software rendering. However because of
this performance difference it is only used when necessary and there
should be no difference for machines supporting OpenGL 3.3+.
The two renderers are largely independent and separated in the
`renderer/text/glsl3` and `renderer/text/gles2` modules. Separate
shaders are also required for text rendering. The rectangle rendering
for underlines and the visual bell works identically for both versions,
but does have some version-specific shader code.
Fixes#128.
Co-authored-by: Christian Duerr <contact@christianduerr.com>
This commit finishes the effort from a64553b to avoid reloading font
twice during startup, since the original issue is with getting font
metrics without building the glyph cache.
Instead of creating a `Rasterizer` to guess the window dimensions,
dropping it and then creating a new one for the glyph cache, reuse the
same `Rasterizer`.
This prevents the font from being loaded twice during startup.
If font size changes however the cells stay the same the terminal won't
be damaged, since it wasn't resized, however the visual change happened,
thus the entire screen should be damaged.
We've been testing out the pre-release feature of crates.io and there
haven't been any major issues. So this updates the documentation to
recommend publishing to crates.io even before the official release.
This adds a codesign step to our makefile build which should
automatically sign the application and resolve the issues on M1 macs
where Alacritty constantly requests permissions.
Since self-signing does still seem to fix this issue after sharing the
resulting `Alacritty.app` with other M1 macs, the binary produced by our
release build should also be fixed automatically.
Fixes#5840.
This allows compositors to only process damaged (that is, updated)
regions of our window buffer, which for larger window sizes (think 4k)
should significantly reduce compositing workload under compositors that
support/honor it, which is good for performance, battery life and lower
latency over remote connections like VNC.
On Wayland, clients are expected to always report correct damage, so
this makes us a good citizen there. It can also aid remote desktop
(waypipe, rdp, vnc, ...) and other types of screencopy by having damage
bubble up correctly.
Fixes#3186.
While using underline thickness could sound logical to draw other
lines most fonts don't make underlines thick compared to cell bounding
box if you increase font size. So instead we're using cell width to
scale builtin font nicely.
This commit also adjusts arcs drawing and alignment.
Fixes#5826.
Fixes#5821.
Alacritty's website (https://alacritty.org) is now advertised publicly
on the GitHub page. Since that includes the old 0.3.0 and 0.5.0
changelogs, it's not necessary to point them out separately anymore.
In 7398e9f a regression was introduced which causes Alacritty to crash
on startup since wayland has a keyboard repeat rate thread started
before our logger is initialized.
Since the latest version of time was rather inconvenient to use anyway
and there is no nice solution for this issue other than downgrading the
`time` version again, the time since startup is now logged instead of
the local time.
This should still provide all the relevant information, while getting
rid of an unnecessary dependency. While it would be possible to also
print the delta between log messages, this can be trivially computed so
it has been omitted to skip adding another `Mutex` to the `Logger`
struct.
Since the CI machine is amd64, it is not possible to just execute the
aarch64 binary to test the application. So instead of running `cargo
test`, we just use `cargo build` instead.
Building on the non-release CI instead of just checking also allows us
to make sure linking works properly before a release fails due to it.
Due to unsoundness issues (c.f., time-rs/time#380 and time-rs/time#293),
determining the local timezone can only happen while single-threaded.
Determine the timezone early in startup and apply the offset to the UTC
timestamp before formatting.
When 'ExpandSelection' binding was added only default binding for
RightClick was added, however to expand block selection holding control
when doing a click is required, so this commit adds a binding for
'RMB + Control'.
This resolves an issue where the regex search could loop indefinitely
when the end point was defined in a location containing a fullwidth
character, thus skipping over the end before termination.
Fixes#5753.
This fixes a regression introduced in 8e584099, where block selections
containing the last cell would have the trailing newline stripped and be
joined into one long line on copy.